How does memory interface work?

[Thomas_K]

Looking at the leaked specs of the AMD Fury X it shows a 4096 bit memory interface. I'm not familiar with what memory interface actually does. But considering this is about 4x faster than previous cards what should we be seeing performance wise?

[Toddwjp]

its how much data can pass across the bus in one cycle, so a 4096 bit bus lets 4096 bits across in one cycle, multiply this by the memory clock speed speed and you get the total data that can be transferred per second, usually measured in gbits

[Thomas_K]

so it should allow better FPS at high resolutions?

[TheRetiredSlave]

its how much data can pass across the bus in one cycle, so a 4096 bit bus lets 4096 bits across in one cycle, multiply this by the memory clock speed speed and you get the total data that can be transferred per second, usually measured in gbits

 

To add to this: We have absolutely no idea how this will affect performance.. Generally speaking, having more memory bandwidth will can (?) improve performance at very high resolutions, which need lots of bandwidth to move the massive amounts of data around.

 

AFAIK, the upcoming Fury X uses a very different memory architecture than anything we've seen in the past. Unless I've missed something, we basically have no freakin' clue how it will perform at varying resolutions. 

 

I would assume the performance probably isn't going to be revolutionary. In the past, we've seen that overclocking memory speeds have had less of a performance impact than overclocking the core. This would lead me to believe that memory bandwidth is probably less important.. That being said, since we're talking about something that's insanely higher than anything else, we don't really know what to expect.

 

Edit:

so it should allow better FPS at high resolutions?

 

Maybe..? This would only really be the case if the only reason framerate was limited was due to a lack of memory bandwidth. If memory bandwidth wasn't the primary bottleneck, I would assume that performance probably won't be very different than anything else.

[Toddwjp]

so it should allow better FPS at high resolutions?

 

 

To add to this: We have absolutely no idea how this will affect performance.. Generally speaking, having more memory bandwidth will improve performance at very high resolutions, which need lots of bandwidth to move the massive amounts of data around.

 

AFAIK, the upcoming Fury X uses a very different memory architecture than anything we've seen in the past. Unless I've missed something, we basically have no freakin' clue how it will perform at varying resolutions. 

 

I would assume the performance probably isn't going to be revolutionary. In the past, we've seen that overclocking memory speeds have had less of a performance impact than overclocking the core. This would lead me to believe that memory bandwidth is probably less important.. That being said, since we're talking about something that's insanely higher than anything else, we don't really know what to expect.

 

Edit:

 

 

Maybe..? This would only really be the case if the only reason framerate was limited was due to a lack of memory bandwidth. If memory bandwidth wasn't the primary bottleneck, I would assume that performance probably won't be very different than anything else.

 

pretty much, its very rare for memory to bottleneck fps currently, since at the required resolution and fps the gpu horsepower is usually what holds us back, there are situations where the memory bus is the limiting factor, however these are not gaming related, usually workstation task's.